Prompt Card Library

This page contains prompt cards — ready-to-use prompts that you can paste directly into a coding agent (such as GitHub Copilot, Claude Code, or similar tools) to generate analytics outputs from Viva Insights data.

Each prompt card includes the purpose, required inputs, assumptions, the full prompt text, adaptation notes, and common failure modes.

How to use a prompt card

  1. Prepare your data. Export the required query from the Viva Insights Analyst portal.
  2. Open your coding agent. Launch GitHub Copilot, Claude Code, or your preferred AI assistant.
  3. Copy the prompt. Find the relevant card below and copy the full text from the Prompt section.
  4. Paste and run. Paste the prompt into the agent. Point it at your data file.
  5. Review and adapt. Check the output against the Common failure modes section. Use the Adaptation notes to customize.

Tip: You can prepend context to any prompt. For example: “My CSV is at ./data/person-query.csv. The Organization column is called Org.” followed by the full prompt text.

Available prompts

Copilot Adoption

Prompt Card Description
Dashboard Overview Generate a comprehensive static HTML dashboard showing Copilot adoption trends, usage patterns, and organizational breakdowns.
Executive Summary Produce a concise executive memo summarizing Copilot adoption metrics for VP/C-suite audiences.
Segmentation & Churn Classify users into usage segments, track transitions, and calculate churn rates.
ROI Estimation Estimate return on investment for Copilot by quantifying time savings and license costs.
Executive PowerPoint Deck Generate an exec-ready 10–15 page PowerPoint deck with editable native charts.
Copilot Causal Toolkit Run a causal inference analysis using the Copilot Causal Toolkit, then interpret results for senior leadership. Two prompts: one to run the analysis, one to interpret outputs.

Purview Augmentation

Prompt Card Description
Agent Usage Analysis Analyze Copilot agent and extension usage patterns from Purview audit logs.
Audit Log Parsing Parse and clean raw Purview audit log exports into analysis-ready flat tables.

Tips for adapting prompts

  • Column names vary between tenants. Always verify your actual column names against what the prompt expects. Prepend a note like “In my data, the Organization column is called Org.”
  • Granularity matters. Most prompts assume person-week data. If your export is person-day, instruct the agent to aggregate first.
  • Privacy thresholds. For smaller organizations, add a note requesting minimum group sizes (e.g., suppress segments with fewer than 5 people).
  • Language preference. Prompts support both R and Python. Choose whichever is already installed in your environment to minimize setup. If you have a strong preference, add “Use R” or “Use Python” before the prompt.
  • Intermediary files. Prompts that produce HTML output instruct the agent to create an RMarkdown (.Rmd) or Jupyter notebook (.ipynb) first, then export to HTML. Keep these intermediary files — they make troubleshooting and iteration much easier.
  • Package availability. Prompts reference the vivainsights R package and vivainsights Python package. Install them beforehand.